What is the “MP” denoiser?
Multipass - This is the denoiser option which worked in older standalone denoisers and yet do not. The blurry effect in recent denoiser is due to limitation, which only does denoise on the beauty pass.
Go back to 3.7, older denoiser and compare rgb denosing vs multipass. The rgb method is the same as in V5. I’ve described it at least once on the forum.
I assure you that’s not the case, even the the RGB denoising mode on 3.7 was a lot better than what we have now (the classic living posted before was done in 3.7 BF + RGB denoiser and as you can see textures are not blurred). The multipass denoiser was even better but in some circumstances the RGB mode was superior.
My guess is that actual denoiser do not read the needed channels for some reasons and so it simply blur the output without taking in to account textures, normals, noise levels etc.
According to Vlado ( https://forums.chaosgroup.com/forum/…-store-concept ) you can get the multipass denoiser even on 5 by manually add the denoiser to the channels and rebuilding them in the VFB, I’m a bit busy right now but if you have time you can check if that works for you;)
@shaio @vlado it would be surely useful to have some input from developers 
Could you please make someday some comparisons? I never used rgb denoise, because it always would loose details. But I do believe, that current rgb can be even worse. Last time I checked it was similar.
Guys could we pls stay at the topic? I wanna know if I have to believe that Rendertimes are now set to slow motion mode because of the nature of Bruteforce/Light Cache?
- So we are not able to chose IR/LC anymore which in my opinion is so ultrafast for Interiors especially?
- Graphicscardrendering seems to be drastically faster. So I have to throw my Threadrippers out of the window now?
- Why was IR/LC Solution cut? The simple answer: “Because its outdated” doesnt count for me. Its extreme fast and when you tweak it, you get a lot Out of it. So we basically now stuck in “Iam a customer and to lazy or not educated enough to set up IR/LC so I just chose 3 options, bruteforcing my way through, do not have to think but Rendertimes are then 20x slower”. What is this approach? This is a slap in the face of every professional.
- Give us pls IR/LC back…
When I’ll have some spare time I’ll make some comparison:) If you check some of my works from last year in the gallery, you can see the results, classic projects are mostly denoised using the RGB mode.
In my experience the multipass mode was a bit better on details but worse on other parts and it tend to produce artifacts, so in time I’ve started to use the RGB mode more and more, and to be honest I’vent missed the multipass mode that much. Personally I’m ok with the RGB mode only on V5, but it must be fixed, in current state as said is barely usable..
Render time with BR are not 20x slower, overall I would say that IM are about twice as fast, if you take in to account the time needed for the setup and the possible issues/lower detail quality I would say that a little time penalty doesn’t matter that much. That’s why nobody develop biased solution anymore, after all most modern renderer never had a biased system, are not any faster than Vray BR and users are working with them just fine.
When 3.x version was released I was thinking just like you (I even say that clearly on a forum, I’ll try to find the topic if you don’t believe me ), back then for me BR was a no go because of the noise/render time and especially for my classic works the IM was the only way. As soon as denoiser was added in version 3.4 I totally change my mind, it turn out that using BR I was more productive, the GI quality was improved and the denoiser take care of the noise. Once you embrace the BR you never go back to IM.***
GPU rendering is not drastically faster, in some circumstances it is surely faster, in other CPU may still be faster and do not have to deal with GPU issues. In our internal tests while developing our shaders for the same render a 32core 3970X TR took about 1’, a 2080Ti 50’'. Each system have some advantage, personally for my works I would still choose the Threadripper especially now that GPU prices are highly overpriced;)
***if you do have a capable denoiser!!!
Dont know what you are talking about @sirio76 . Iam testing with my Threadripper against my shitty old 1070ti and it loses by a BIG MARGIN.
AA is the same. Settings are the same.
As said speed gain is scene dependent, I’ve just reported my personal experience;) Just be aware that render settings are not comparable between CPU and GPU, they are two different engine and you should not expect same results using similar settings.
Hope @shaio will respond some time. Also I have transparency Issues when rendering the same scene with GPU. When rendering with CPU I can see through the Window onto the roof. But when rendering the same sfene with GPU I can clearly see everything through the window when Light Cache is built but when rendering Starts, something weird is going on. I cant even explain what happens. I will post an image later. Its really annoying to deal with all that weird stuff… Stealing all my time. I would appreciate if Chaosgroup would me more transparent about things.
Also waiting so Long for responses. Dont know exactly why. When I post something in the Redshiftforum its answered after 2-3 hours or so. Also they give me full advice and analyze things.
Maybe V-Ray for c4D is still not being taken seriously. Its a real pain. Even Chaosgroup must SEE that so many people are unhappy. It would just Help a lot and give hope if they just be more transparent.
Its like they are imitating Nintendo. Being full silent and at the end you get what you get but not what you expect/want.
I’m on Max but my guess is that it is the ‘Trace Depth’ (defaults to 5) levels in GPU performance settings - set it to a higher number and that may fix the ‘weird’ window situation…
Also I have transparency Issues when rendering the same scene with GPU. When rendering with CPU I can see through the Window onto the roof. But when rendering the same sfene with GPU I can clearly see everything through the window when Light Cache is built but when rendering Starts, something weird is going on. I cant even explain what happens.
V-Ray and V-Ray GPU are two separate renderers, and they should be accepted as such. Whenever you start a project and chose your render engine, you should stick with it. A lot of the algorithms that are calculating different aspects of the scene are vastly different for GPU devices compared to CPU devices because of the different architecture therefore in some case one of the engines will be fast, where in others not so fast. I can give you as an example the Min/Max subdivisions of the Bucket sampler. If you use the Min=1 and Max=24 values on V-Ray GPU the same way as you use them with V-Ray, you won’t get the same noise amount as you would get with V-Ray only. And this isn’t because one is of the engines is faster than the other. It is because they’re using different approaches.
Nowadays, V-Ray GPU supports a large portion of the available features for V-Ray, but this doesn’t make them and won’t make them the same render engines.
I’m starting to think more and more that we should really separate the two engines fur our future releases.
Also waiting so Long for responses. Dont know exactly why. When I post something in the Redshiftforum its answered after 2-3 hours or so. Also they give me full advice and analyze things.
Maybe V-Ray for c4D is still not being taken seriously. Its a real pain. Even Chaosgroup must SEE that so many people are unhappy. It would just Help a lot and give hope if they just be more transparent.
I’m sorry to hear you still think that we’re not taking everything seriously, even after the poll that you initiated and after seeing how much of the features from the same poll went in the Update 1 release of V-Ray 5. We will of course continue to work on improving V-Ray and do our best as much as we can.
As for the scene, I’ve started to look in to it and see what could be the problem for the lost transparency on glass with V-Ray GPU. Checking the rendertimes would take some time, since I need to do a few renderings until I see if there is something that could be optimized for the materials or anything else in the scene.
Also I have transparency Issues when rendering the same scene with GPU.
I’ve checked the scene about the transparency and as suspected you’ve reached the internal limit of maximum depth for refraction, which for V-Ray GPU is 16. I’ve counted around 24 polygons with glass material that the rays should pass through and after the 16th it starts to render black. It just happens that we’re currently working on overall improvements on V-Ray GPUs core, and they will lead to the removal of this limitation.
Ah thank you very much @shaio . That explains what is happening here. But this helps me to figure out a workaround.
And to be clear: Iam just a stubborn guy. I sometimes go overboard with my expectations. Iam very happy how development is going right now! Thank you very much that you looked into the scene.